Speak.
Your Mac types.

Voice input that lives in your menu bar. Hold a key, speak naturally — polished text appears at your cursor, entirely on-device.

Click play to see it transcribe.
Screen context Fake email thumbnail showing flight confirmation

Hold fn

Press and hold your configured key — fn, ⌃, ⇧, or ⌥. Recording starts instantly.

Speak

Talk like you would to a person. Audio never leaves your Mac.

Release

Let go. Polished text appears at your cursor, in any app.

Nothing leaves your Mac.

WhisperKit transcribes locally. MLX-Qwen refines locally. Audio is never uploaded, never stored.

Under the hood

WhisperKit
Offline speech recognition
MLX-Swift-LM
On-device LLM inference
Swift 6
Native, no cross-platform wrappers
ScreenCaptureKit
Screen context via OCR
AVAudioEngine
Low-latency mic capture
Vision
On-device OCR for homophone correction

Runs on macOS 26+ · Apple Silicon only · MIT licensed

Ready to try?

Download for macOS
xattr -cr /Applications/OpenType.app copy

Run once if macOS shows a Gatekeeper prompt.

Works with OpenAI, Claude, Gemini, OpenRouter, SiliconFlow, Doubao, Bailian, MiniMax — or any OpenAI-compatible endpoint.